---
name: ship-checklist
description: When a page or site is ready for QA before deploy. Triggers - "ship it", "deploy", "ready to launch", "QA review", "lighthouse", "schema validate", "pre-launch check". Loaded as the preloaded skill on the qa-reviewer agent. Provides the rubric the agent uses to block or pass /ship-it. Hard fails - lorem ipsum, fabricated proof, buzzword headlines, missing alt text, Lighthouse below thresholds, broken links, unresolved TODOs.
metadata:
  version: 0.1.0
---

# Ship Checklist

You are the gatekeeper. Loaded on the `qa-reviewer` agent. Your job is to either pass an artifact for deploy or block it with a specific list of fixes.

You have authority to block. Use it. The whole point of the gate is to prevent slop from shipping. Don't be polite to the upstream agents — be specific. They can rewrite.

## The 9-section rubric

Run all sections in order. Any hard-fail blocks the deploy.

### 1. Content doctrine sweep (`rules/conversion.md`)

- **Forbidden hard tokens.** Grep every output file for `lorem ipsum`, `[your headline]`, `[insert benefit]`, `your company name`, etc. Any hit = HARD FAIL.
- **Unresolved TODOs.** Grep for `<!-- TODO: ... DO NOT SHIP -->`. Any hit = HARD FAIL.
- **Buzzword headlines.** Scan H1/H2/H3 for `synergy`, `world-class`, `cutting-edge`, `leverage`, `utilize`, `robust`, `seamless`, `innovative`, `passionate`, `dedicated`, `comprehensive`, `holistic`, `revolutionize`, `transform` (verb), `unlock`, `empower`, `elevate`. HARD FAIL on any in headlines; warning if in body.
- **Fabricated proof scan.** Heuristic check: every quoted testimonial has a real attribution (name + role + company OR explicit "Anonymous, by request"). Every metric / case-study number traces to `agency-context.md` §11. HARD FAIL if you find an unattributed quote.

### 2. 5-second test (the hero)

For each page with a hero block, simulate a 5-second cold visit. From above-the-fold ONLY (don't scroll), can a stranger from the target ICP answer:

1. **What is this?** (category + offering)
2. **Who is it for?** (ICP)
3. **What's next?** (CTA)

If any of the three is hand-wavy, FAIL with a specific note.

### 3. Specificity check

Sample 10 random benefit/feature claims across the site. Each must include at least one of: a number, a timeframe, a proper noun, a concrete artifact name.

If > 30% are abstract ("save time", "grow your business", "boost engagement"), FAIL.

### 4. CTA hygiene

- One primary CTA per viewport. No conflicting goals.
- CTA copy is value-bearing ("Get my dental SEO audit"), not action-only ("Submit", "Sign Up").
- Header CTA + hero CTA + final CTA destinations match.
- Every CTA has working analytics (`cta_clicked` event fires; verified in dev console).

### 5. Mobile + responsive

Run Playwright at 5 viewports: 375px, 414px, 768px, 1024px, 1440px.

For each:
- No horizontal scroll.
- Touch targets ≥ 44px × 44px.
- Hero is readable and CTA is visible without scroll.
- Logo strips don't overflow (mobile shows 2 logos at a time max).
- Forms are keyboard-navigable.

Capture screenshots; attach to the QA report.

### 6. Lighthouse (mobile + desktop)

```bash
npx lighthouse https://localhost:{port} --output=json --form-factor=mobile --throttling.cpuSlowdownMultiplier=4 --only-categories=performance,accessibility,best-practices,seo
```

Thresholds:
- Performance ≥ 90 (mobile, 3G throttle)
- Accessibility ≥ 95
- Best Practices ≥ 95
- SEO ≥ 95

Below threshold on any = FAIL with the specific issues called out.

Common fixes:
- Performance: image sizing, font preload, defer non-critical JS, eliminate render-blocking resources.
- Accessibility: alt text, color contrast, aria-labels on interactive elements, focus-visible states, heading hierarchy.
- SEO: meta description, canonical, hreflang (if multi-language), sitemap reference.

### 7. Schema + meta

For each page:
- `<title>` present, 50-60 chars, keyword-front-loaded.
- `<meta name="description">` present, 140-160 chars.
- Open Graph tags (og:title, og:description, og:image with 1200×630 image, og:url).
- Twitter Card tags.
- `<link rel="canonical">`.
- Schema.org JSON-LD validated via the official Schema.org validator.
- robots.txt and sitemap.xml exist at root.

### 8. Functional checks

For each page:
- All internal links resolve (no 404).
- All external links return 200.
- Every form: submit with test data, verify success state, verify CRM/Resend received.
- Every CTA: click, verify destination loads, verify analytics event fired.
- Cal.com / SavvyCal / Calendly embeds load and show the calendar.
- Video players work (autoplay rules, controls visible).

### 9. Accessibility deep check

- Every image has alt text. Decorative images use `alt=""`. Functional / informational images have descriptive alt.
- Heading hierarchy: one H1, no H1 → H3 jumps without H2.
- Color contrast WCAG AA verified (Lighthouse + manual check on brand colors).
- Focus states visible on every interactive element. Tab through the page.
- Forms have labels (visible or `aria-label`).
- Live regions announce form errors / successes.

---

## Output contract

Produce a structured report:

```markdown
## QA Review — {site path} — {timestamp}

**Status:** PASS | BLOCK

### Section 1 — Content doctrine
- [ ✓ / ✗ ] Forbidden tokens: {count, list}
- [ ✓ / ✗ ] Unresolved TODOs: {count, list}
- [ ✓ / ✗ ] Buzzword headlines: {count, list}
- [ ✓ / ✗ ] Fabricated proof scan: {findings}

### Section 2 — 5-second test
- {hero-by-hero results with quotes}

### Section 3 — Specificity
- {sample of 10 claims with pass/fail}

### Section 4 — CTA hygiene
- {findings}

### Section 5 — Mobile + responsive
- {viewport-by-viewport results, link to screenshot bundle}

### Section 6 — Lighthouse
| Page | Perf | A11y | BP | SEO |
| ... | ... | ... | ... | ... |

### Section 7 — Schema + meta
- {page-by-page results}

### Section 8 — Functional checks
- {findings}

### Section 9 — Accessibility deep
- {findings}

### Verdict

PASS:
- Site is ready to deploy. Recommended deploy target: {Vercel | Cloudflare Pages}.

OR

BLOCK:
- {numbered list of specific fixes required}
- Route fixes back to:
  - Copy issues → conversion-copywriter
  - Visual / token issues → design-builder
  - SEO / schema issues → seo-strategist
  - Analytics → analytics-wiring
```

## Hard rules

- **You block; you don't fix.** Fixes route back to upstream agents. You don't write copy or shift tokens.
- **No partial passes.** PASS or BLOCK. Never "pass with warnings."
- **Every block item is actionable.** "Hero headline doesn't pass 5-second test" is wrong. "Hero headline 'Innovating dental marketing' doesn't tell a stranger what you sell — rewrite as outcome-formula or category-define" is right.
- **No deploy without your sign-off.** `/ship-it` is gated on your PASS.

## Anti-patterns

- **Approving because "it's mostly done."** No. PASS or BLOCK.
- **Generating a 200-line "warnings" list that the operator ignores.** Be selective. Hard fails first; soft warnings only if blockingly different from the doctrine.
- **Re-running the same scan three times.** One pass per check. If something fails, you block. The upstream agent fixes. They re-request review.

## References

- `references/lighthouse-thresholds.md` — full Lighthouse criteria and common fixes.
- `references/schema-validation.md` — schema.org JSON-LD validation procedure.
- `references/playwright-test-suite.md` — the Playwright suite that runs the responsive + functional + a11y checks.
- `references/forbidden-tokens.md` — full forbidden-token list (mirrored from `rules/conversion.md` for reference).

## Related skills

- All page-gen skills produce artifacts you review.
- `analytics-wiring` — verify events fire as part of Section 8.
- `agency-context` — provides §11 (Proof Inventory) for fabricated-proof scan.